Do languages ever shift from analytic to synthetic?

June 12, 2009 3:45 PM

Calling all historical linguists: do languages ever shift from being analytic to synthetic rather than vice versa?

This is a very non-scientific treatment of the issue--and the presumption that it's probably wrong is why I'm here--but it seems like whenever I read about the history of a particular language, the language has 'lost' a lot of features compared to its ancestral forms. I know that these forms of expression are rarely 'lost' in the sense of being completely gone, rather replaced (prepositions taking over the roles of the declensions in the Latin -> Romance transition, for example) but it seems like complicated inflectional morphology has a tendency to be lost at some point in favor of an analytic, word order- or adposition-based system of expressing these relationships between words. I know that in certain cases you have languages gaining pronouns from other kinds of linguistic expressions ('usted' and 'você' in Spanish and Portuguese, 'a gente' replacing 'nós' in Brazilian Portuguese, etc.), but it seems like the farther back you go in tracing a language's evolution, the more likely you are to find a language with more morphologically-complex forms, a complicated case system with declensions, etc. (This question was sparked by reading about the Iranian elections today, which led to reading about Persian and encountering yet another case where the earlier forms were more morphologically complex than the modern form.)

So, why does it seem like complex inflection is generally lost? I understand that sound change is a volatile process and sometimes word forms fail to maintain phonetically distinctive features, but with that being the case, does anyone have any idea as to how the complex systems they simplified from arose in the first place? Or failing that, are there any examples of a language moving from being primarily analytic to primarily synthetic? The examples I encounter all seem to move almost exclusively in the other direction. Sometimes languages maintain a level of inflection that's higher relative to their sister languages (no one would argue with Spanish being classified as more inflected than English), or they gain new synthetic forms in some isolated cases (the Romance future and conditional) but their morphology is still significantly simplified in comparison with the parent form.

(I know that these examples are all Indo-European languages so I would welcome examples from other language families as well.)

posted by Kosh to writing & language (16 comments total)

First of all, some clarity for other people reading this thread who have no idea what synthetic or analytic languages are, because I had to go look it up. An analytic language conveys meaning primarily through "use of particles and word order rather than by inflection." Synthetic languages contain words which are more likely to share meaning depending on their inflection. This distinction is typically more useful in a comparative context such as 'language X is more synthetic than language Y' than in its ability to put a certain language in a box on one side.

That being said, my guess is that it would be more common for languages to drift toward the analytical side than vice versa because of the ambiguity which reliance on inflection introduces, as well as the fact that it just does not translate cleanly into a phonetic alphabet. The "complex systems they simplified from" were likely due to the fact that those who developed the language did not plan it out ahead of time, and developed ad-hoc inflection strategies to differentiate certain things. Over time, others became frustrated with this lack of clarity, and cleaned things up.

IANAL
posted by sophist at 4:13 PM on June 12

First of all, some clarity for other people reading this thread who have no idea what synthetic or analytic languages are, because I had to go look it up.

Um. May I politely suggest that if you don't know these terms, you shouldn't be answering the question? Though I do not know the answer (historical linguistics is not my specialty), I am a linguist, and these _are_ standard terms in the field.
posted by advil at 4:26 PM on June 12 [5 favorites has favorites]

Just want to point out that there's a slightly off underlying premise here, in that the languages going from synthetic > analytic are becoming simpler. This is not necessarily true. When something is lost in a language, generally something else is gained, and the language itself maintains its overall level of complexity. English is a good example of this, once being highly morphologically complex (more synthetic) and changing over time to result in the more syntactically complex (more analytic) system we use today.
posted by iamkimiam at 4:35 PM on June 12

Just want to point out that there's a slightly off underlying premise here, in that the languages going from synthetic > analytic are becoming simpler

Right. Sorry. I should have been clear in the question that when I make reference to "simplification" what I'm referring to is the replacement (be it partial or full) of morphology-based systems with syntax-based systems, often as a result of sound changes making previously distinct forms identical (and thus necessitating a shift to some other way of disambiguating), not meaning to imply that, say, something like Modern English is somehow a fundamentally 'simpler' language than Old English in terms of its ability to express things just because its morphology is simpler.

(I realize this is a pretty technical question for a limited audience, but I wasn't sure where else to ask it, and I had seen questions of similar complexity when I was browsing the archives, so...)
posted by Kosh at 4:41 PM on June 12

Two links that might make interesting reading until either languagehat or a linguistics grad student arrives to school us all:

• "We have to look at CONSTRUCTIONS, not at LANGUAGES. We then see that there is a universal directionality of change: Analytic constructions always turn into synthetic constructions (unless they die out), and synthetic constructions are always replaced by newly created analytic constructions."
• Grammaticalization
• During this period the system of nominal suffixes (inflection) is still in the making; its development lasts till the late old Hungarian period. [...] Certain postpositions become suffixes. The transitional phase is shown by the development of suffix -re/-re, for example: in the Letter of Foundation of Tihany it is still a postposition: "ohut cuta rea" (óút kútja reá), "azah fehe rea" (aszófő reá), but in the 13th century it is already a suffix: "Balwankure" (Bálványkőre).
• "I just ordered a dictionary of Indo-European roots to play with, and I started wondering -- how did we get inflectional endings? Were they originally postpositions which became attached to the nouns?" "That's the usual hypothesis."

(Note: I can't vouch for the exact accuracy of everything linked above, but I think the general idea — that all languages are a mixture of construction types, and that you have to think about how these constructions themselves might change rather than a switch suddenly flipping from "analytic" to "synthetic" — is sound.)
posted by No-sword at 4:43 PM on June 12

And by two I mean four.
posted by No-sword at 4:43 PM on June 12

On the matter of why: I have no idea. I would suggest that the development of the Romance languages is the best attested and understood example of this phenomenon. It may not even be a unitary phenomenon, in the sense that superficially similar processes and controlled by different factors. An example would be creolization, which always (?) leads to analytical languages, yet certainly isn't a factor in most of the language change you're talking about. In the case of Old English, the loss of vowel distinction seems to be the proximate cause in a move to a more anaytical system, although there is a period where both fixed word order and vowel distinction is present in some texts, and yet both are absent in others.

As to whether they go the other way: kinda. I think the standard textbook explanation (beyond which I'm really not educated), is that agglutinating in isolating languages leads back to morphological complexity.

(PS I could be wrong on all of this, I'm a dabbler only.)
posted by Sova at 4:49 PM on June 12

Page 183 of the Ergativity by Robert W.M. Dixon, suggests that there is a cycle of language change, including a step where languages can go from isolating to agglutinative. I can't think of any specific examples (but like you, I can think of many going the other direction), but that isn't to say it doesn't happen.

I studied Zulu last semester, which is agglutinative (a form of synthetic) and given its 17-19 noun class prefixes, its large phonological inventory, its consistent use of tone and prosody to mark word boundaries and such, I could imagine a scenario where it would naturally become even more agglutinative in order to reinforce the system and strengthen the multi-operational paradigms. I have no idea if that's the case, but it makes sense to me that a language like this could end up going that direction in an organic, natural way.
posted by iamkimiam at 4:50 PM on June 12

Very interesting and helpful stuff so far, thanks. My only formal experience with historical linguistics has been in history of Spanish classes (which is probably obvious from my question... I know more about historical Romance than I do my native language's history), so I don't know much about the field from a broader point of view.

I certainly didn't know that there were people who hypothesized that the PIE inflection originated from fused postpositions... all the stuff I've ever read in my dabbling seems to have been written with this underlying assumption of "PIE was really inflected and its descendants became less and less so as time went on," which always struck me as odd.
posted by Kosh at 4:59 PM on June 12

Just to follow up on Sova's comment about Old English becoming more analytic. Other factors that greatly influenced this change had to do with the fact that by this time period there was already a relatively fixed word order, and so syntax provided a redundancy and relieved pressure on the morphological system to indicate word function. Also, there were no paradigms with maximum differentiation. Ex. The definite adjectives could have had 30 different inflectional endings (3 genders X 2 numbers X 5 cases), but they only had 5 inflectional endings (-an fills in 17 of the 30). The paradigms became less meaningful and eventually collapsed. And to the original point about vowel distinction, heavy stress on syllable roots with light following led to vowel reduction, which encouraged the loss of inflection. Sidenote: this is part of the explanation as to why today we have many 4 letter words containing long/tense vowels and ending with unpronounced e's.

These types of processes are common ways that a language becomes more analytic. Like you, I'm curious to hear how it all goes down the other way. Where are MeFi's historical linguists today?!
posted by iamkimiam at 5:04 PM on June 12

Are you looking only for inflection as complexity or would you count other synthetic features like agglutination? Because I think this happens, like above, someone points out that creolization leads to analytic languages, but if you take English for example during the period after the Norman conquest when there was a heavy French, and consequently Latin, influence in English, English was in the process of losing its inflected forms, but it also gained the ability to derive new latin words through adding all kinds of latinate affixes to latinate stems. I don't know if it counts for your question though because its additional morphological complexity but its certainly of a different sort than inflection.

Isn't there some language change in isolating languages whereby adpositions become affixes, leading to new inflected forms?

Also, I think you could make the argument that metanalysis is an example of adding a construction of a synthetic nature to an isolating language. To use English again, think of all the words that were borrowed into English as nouns from Latin, and then after a long time, were turned into verbs by deciding that various chunks of the Latinate loanwords are actually morphemes you can use to make new words. To pull some examples from Wikipedia:

A huge number of English verbs of Latinate origin appear to have been back-formed from nouns in -tion, -sion and the like: profess, relate, pollute, confect and literally hundreds more are, with rare exceptions, first attested a hundred years or more after the first attestations of the nouns profession, pollution, confection and so on. (They hardly could come directly from Latin; formally they often look like a Latin perfect participle, which would be an odd source for a verb.)

This is still happening in English today, and it seems as if (American) English has recently become more tolerant of it (even if it annoys the crap out of people). Think of all those 'business' words like 'incentivize'. English is normally an isolating language. When we make business words like "monetize" were are making English more synthetic, the older, analytic way of handling these problems would be to just use the nouns 'incentive' and 'money' and combine them with other simpler verbs ('give people an incentive') or find a different verb that means the same thing ('burgle' vs. 'burglarize'). That Latinate prestige halo affect still has a big effect on consultants, doctors, etc.
posted by jeb at 5:20 PM on June 12 [1 favorite has favorites]

I'm just a dabbler in these kinds of things, so I don't know if this observation is of any use.

I remember from high school that modern German has a tendency to create new nouns by concatanating two or more older nouns together, leading to some really titanic jawbreakers in extreme cases. Does this count as "agglutination" and is it an example of a language moving the opposite direction?
posted by Chocolate Pickle at 5:31 PM on June 12

Also, here are some simplified verb examples from Old vs Modern Japanese that might interest you:

verb: OJ -tari form → MJ -ta form, MJ -masu form
/kas.u/ : /kas.i.tari/ → /kas.i.ta/, /kas.i.masu/
/yob.u/ : /yob.i.tari/ → /yoN.da/, /yob.i.masu/
/kak.u/ : /kak.i.tari/ → /kai.ta/, /kak.i.masu/
/i(h).u/ : /ih.i.tari/ → /it.ta/, /i(h).i.masu/

In OJ, the root of the verb always remains unmolested: /kas/, /kak/, /ih/, /yob/. Centuries of language evolution later, things have developed so that the root changes depending on what the suffix is: in the -masu form, it doesn't change, same as OJ, but in the -ta form, it does.

And the manner in which it changes depends on the sounds are involved. /-si/ is unchanged. /-bi/ becomes /-N/. /-ki/ becomes /-i/. /-hi/ becomes /-Q/ (i.e. doubles the consonant immediately after it).

In other words, bits and pieces that did not originally influence each other phonetically have become "inflections" that are "more complex" in the sense that now they do influence each other, and other parts of the word, based on phonetic factors.

I should note that some people actually consider "kaki tari" in OJ two separate words in the same sense that English "did write" are separate words; if you go along with that analysis, then OJ → MJ is itself an example of what you are looking for, I think.
posted by No-sword at 5:34 PM on June 12

That being said, my guess is that it would be more common for languages to drift toward the analytical side than vice versa because of the ambiguity which reliance on inflection introduces, as well as the fact that it just does not translate cleanly into a phonetic alphabet. The "complex systems they simplified from" were likely due to the fact that those who developed the language did not plan it out ahead of time, and developed ad-hoc inflection strategies to differentiate certain things. Over time, others became frustrated with this lack of clarity, and cleaned things up.

Oh. My. God! Those who developed the language did not plan it out ahead of time? Ad-hoc inflection strategies? Frustrated with lack of clarity? Frightening!

The analytical / synthetic thing is really a spectrum. I speak one highly analytical language - English. One highly synthetic language - Hungarian. And I speak a bunch that are in-between the two - Serbo-Croatian, German, French, etc. All of them have their advantages and disadvantages, and earlier posters are completely correct - the "simpler" analytical languages make up for their "simpler" grammar with all sorts of hard-to-understand weirdness.

In some ways Hungarian is a tough language to learn, because you can't bluff your way through it very well. As a fellow student once told me, "It's like you have to know everything before you can say anything!" Anyone who's ever studied Hungarian or Estonian or Finnish or Turkish or Inuit must know that feeling. On the other hand, the clarity's amazing - contrary to sophist's rather odd assertion. One "word" can carry the subject, verb and object!

hallak = I hear you

And it takes something only a little more than a word to express something really "complex":

Ö napszemüveges? = Are you a habitual wearer of sunglasses?

It's not usually necessary - unlike in analytical languages such as English - to use pronouns. Hungarian doesn't really even have a word for "it," in the way that we normally use it. I can think of a million ways that this clarity is obvious, but one of them might be in describing something like the word "stand" in English. By itself, I don't know if that's an abstract noun (like a political position), a physical object (like a newspaper stand), or a verb. If it's a verb, I don't know if it's used as an infinitive, or in conjunction with "we" or "I" or "they," etc. In Hungarian, each of those usages would exist as a separate and unique word. That's clarity, baby!

The analytic nature of English made it difficult for me to learn. French has nice, one word verbs for things such as "cause," "accomplish," "publish" and "raise." Obviously, English has those verbs, too (mostly taken from the Normans.) But with English, you've got the commonly-used synonyms "bring on," "bring off," "bring out" and "bring up," which use two words - a verb and a preposition - to express something that doesn't make obvious sense to a non-native speaker. One verb might have more than a dozen different (and totally unrelated) meanings because of the many prepositions it can be combined with. This isn't clarity to me! I have a friend who's learning English and is obsessed with soul. She loves James Brown, and knows enough English to understand the odd phrase, "get up," which means, of course, "stand" or "arise." But then what is the meaning of the funky command "Get on up?" I sort of have a sense about it, but if you ask me to succinctly define the "on" in that phrase, I cannot. English has a lot of these small-word idiomatic phrases that are very, very hard to learn.

I can only say this from my own experiences tackling languages which were both more and less synthetic and analytical than my native tongue . . . but although it may be easier to study an analytical language for a few days and make some kind of sense in it, in the long run, the analytical ones are much less clear and obvious.

All of that aside, I think there must be some small instances when a language became slightly more synthetic. English uses a lot of prefixes that don't mean a lot on their own - anti-, omni-, -trans, and so on. These are prefixes that attach to words to alter their meaning, and these are prefixes which came (mostly) from Greek and Latin roots, presumably through the Normans. English has done the same with suffixes. Offhand, I can think of "-gate" to connote scandal, which is obviously recent, and the -nik in a word such as "beatnik," also pretty recent. Although there were precursors to this already in Old English, the language did adopt these synthetic elements even as the language became increasingly analytical, and I'll hazard a guess here that the number of these affixes used this way grew more quickly than original ones were lost - hence, a slight move to syntheticism, if you want to look at it that way - even if most other aspects of grammar moved in the opposite direction.

I'm a fan of language, but not degreed scholar, so I'd welcome any addenda or corrections.
posted by Dee Xtrovert at 6:13 PM on June 12 [8 favorites has favorites]

Guy Deutscher's The Unfolding of Language devotes a whole chapter to the evolution of the Semitic verb over time, from a simple to a very complex set of conjugations. If I understand him correctly, this happened as successive generations perceived, extended and built on patterns that they (mis)perceived in the language - so-called reanalysis. Other chapters provide examples of this kind of process in many different languages of different types.

As a strictly amateur but very keen student of linguistics I found this book fascinating, and people who ought to know better than me seem to rate it as a text for the lay-person. Check it out.
posted by i_am_joe's_spleen at 8:52 PM on June 12

Damn, I'm about to dash off to see my grandson play T-Ball, but I wanted to at least stick my head in and recommend the Guy Deutscher book i_am_joe's_spleen just mentioned; it has excellent and understandable explanations of this stuff. As a general point, yes, it goes both ways, and if you think about it, it logically has to, since language has been around for a long, long time.
posted by languagehat at 7:35 AM on June 13